ARTICLE

https://doi.org/10.1038/s41467-020-15598-x OPEN XPO5 promotes primary miRNA processing independently of RanGTP

Jingjing Wang 1, Jerome E. Lee1,3, Kent Riemondy1,4, Yang Yu2, Steven M. Marquez1, Eric C. Lai 2 & ✉ Rui Yi 1

XPO5 mediates nuclear export of miRNA precursors in a RanGTP-dependent manner. However, XPO5-associated RNA species have not been determined globally and it is unclear

1234567890():,; whether XPO5 has any additional functions other than nuclear export. Here we show XPO5 pervasively binds to double-stranded RNA regions found in some clustered primary miRNA precursors and many cellular RNAs. Surprisingly, the binding of XPO5 to pri-miRNAs such as mir-17~92 and mir-15b~16-2 and highly structured RNAs such as vault RNAs is RanGTP- independent. Importantly, XPO5 enhances the processing efficiency of pri-mir-19a and mir- 15b~16-2 by the DROSHA/DGCR8 microprocessor. Genetic deletion of XPO5 compromises the biogenesis of most miRNAs and leads to severe defects during mouse embryonic development and skin morphogenesis. This study reveals an unexpected function of XPO5 for recognizing and facilitating the nuclear cleavage of clustered pri-miRNAs, identifies numerous cellular RNAs bound by XPO5, and demonstrates physiological functions of XPO5 in mouse development.

1 Department of Molecular, Cellular, and Developmental Biology, University of Colorado Boulder, Boulder, CO, USA. 2 Department of Developmental Biology, Memorial Sloan Kettering Cancer Center, 1275 York Avenue, Box 252, New York, NY, USA. 3Present address: ArcherDX, Boulder, CO, USA. 4Present address: ✉ RNA Bioscience Initiative, University of Colorado School of Medicine, Aurora, CO, USA. email: [email protected]

NATURE COMMUNICATIONS | (2020) 11:1845 | https://doi.org/10.1038/s41467-020-15598-x | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15598-x

PO5 is a karyopherin protein1 that mediates nuclear To test the function of XPO5 in mouse development, we show Xexport of miRNA hairpin precursors (pre-miRNAs)2–4. that constitutive deletion of XPO5 leads to early embryonic XPO5 binds to hairpin RNAs with 2–8nt 3′ overhang lethality and failed gastrulation at approximately embryonic day including pre-miRNA and minihelix viral RNA in a RanGTP- 7.5. For tissue morphogenesis, we find that conditional knockout dependent manner2–5. The formation of XPO5:pre-miRNA: (cKO) of XPO5 in the epithelial cells of the skin shows similar but RanGTP nuclear export complex was confirmed by X-ray crys- slightly different defects than those observed in Dicer1 or Dgcr8 tallography6, in which RanGTP binding triggers a conformational cKO17–19. Quantitative measurement of miRNA biogenesis in the change of XPO5 and several XPO5 residues in the HEAT repeats skin shows ~90% global loss for most canonical miRNAs except form hydrogen bonds with 5′ and 3′ ends including the 2-nt 3′ for a few Drosha/Dgcr8-independent miRNAs such as mir-320 overhang of pre-miRNAs. Functional studies showed a require- and mir-48419. Taken together, this work provides insights into ment of XPO5 to transport pre-miRNAs from the nucleus to the the function of XPO5 in miRNA biogenesis that is independent of cytoplasm2–4, providing the evidence that XPO5 is an essential RanGTP binding and nuclear export, and identifies numerous component of miRNA biogenesis7. However, XPO5-independent cellular RNAs as XPO5 binding partners. These findings establish pre-miRNA export pathway was also reported8,9. Notably, 7- a molecular basis to further investigate the regulatory roles of methylguanosine-capped pre-miRNAs whose biogenesis is inde- XPO5 in miRNA biogenesis and the metabolism of other pendent of the DROSHA/DGCR8 microprocessor are exported cellular RNAs. via the PHAX-XPO1 pathway9. In a recent study, deletion of XPO5 in human HCT116 cells only mildly affected the biogenesis of most miRNAs, suggesting that XPO5 is sufficient but not Results required for miRNA biogenesis10. However, in a CRISPR-Cas9 XPO5 associates with primary and pre-miRNA precursors.UV screen for genes that are critical for mir-19-mediated silencing crosslinking followed by immunoprecipitation and sequencing fi 11 has been successfully used to identify highly structured miRNA also in human HCT116 cells, XPO5 was identi ed as essential . – In addition to these different reports, the lack of genetically precursors that are bound by DROSHA, DGCR8, and DICER120 engineered mouse models of XPO5, unlike other core compo- 22. To globally identify XPO5-associated RNA species in human nents of the miRNA pathway such as Drosha, Dgcr8, Dicer1,or cells, we applied a high-throughput sequencing of RNA isolated Ago genes, has hindered the understanding of XPO5 functions. by crosslinking immunoprecipitation (HITS-CLIP) approach to Overall, the in vivo requirements of XPO5 for global miRNA map RNA fragments bound by endogenous XPO5 in biogenesis, mammalian development, and tissue formation HEK293T cells (Fig. 1a). In the absence of XPO5 antibody, we did remain to be determined. not recover any RNA fragments upon immunoprecipitation XPO5 is a highly expressed protein among core components of (Fig. 1a). Five independent HITS-CLIP libraries were generated miRNA biogenesis in both mouse and human cells12. However, it and they displayed generally consistent profiles of XPO5- is more sensitive to saturation caused by the expression of short associated RNAs. To increase stringency, we required the detec- hairpin RNA and highly structured viral RNA than other com- tion of any RNA fragments in at least three out of the five ponents such as Drosha and Dicer113,14. Although individual libraries for further analysis. In total, 64.2 million reads were RNA species such as pre-miRNA2–4, some cellular tRNAs15 and generated and mapped to 83,169 genomic peaks (see “Methods”). minihelix viral RNAs such as adenovirus VA1 RNA5,14 have been Because XPO5 is known to interact with pre-miRNA hairpin and identified as XPO5 cargoes for nuclear export, XPO5-associated mediates their nuclear export2–4,wefirst determined whether cellular RNAs have not been determined at the genomic scale. It pre-miRNA hairpins were recovered in our datasets. Indeed, all is unclear how many pre-miRNAs directly interact with XPO5 37 pre-miRNAs of the most highly expressed miRNAs, judging and require XPO5 for their biogenesis. Furthermore, since the from miRNA reads number (>1000 reads) from a quantitative discovery of XPO5 as the nuclear export factor for pre-miRNA smRNA-seq in the HEK293T cells (Supplementary Data 1), were hairpin2–4,6, much attention to XPO5 has been drawn to its detected in XPO5 HITS-CLIP (Fig. 1b). Overall, 96 out of the top nuclear export functions. However, XPO5 was originally identi- 107 miRNAs, whose read number was >200 in the smRNA-seq, fied as a binding protein of double-stranded RNA (dsRNA) were also detected in our XPO5 HITS-CLIP. Furthermore, XPO5- binding proteins such as ILF3, PKR, and Staufen16. Although the associated miRNA sequences generally spanned the left arm, the direct binding of RNA species including pre-miRNAs, tRNA and top loop, and the right arm regions, reflecting the association of VA1 RNA to XPO5 has since been well documented2–6,14,15,it XPO5 with pre-miRNA hairpins prior to DICER1 cleavage. Only remains unknown whether other cellular RNAs with double- ~20% of XPO5 HITS-CLIP reads were mapped to either arm stranded regions can bind to XPO5. In addition, all known (Fig. 1c), which generally represents mature miRNAs. In com- RNA substrates of XPO5 bind to XPO5 in a RanGTP-dependent parison, >98% of mature miRNA reads recovered from smRNA- manner, which is a key feature of karyopherin-mediated seq were mapped to the left or the right arm as expected (Fig. 1c). nuclear export1. As a result, whether XPO5 plays any roles in To compare the profile of XPO5-associated miRNA sequences cellular RNA metabolism other than nuclear export is an open with other essential components of the miRNA biogenesis question. pathway including DROSHA, DGCR8, DICER1, and AGO To address these questions, we first perform HITS-CLIP ana- proteins, we downloaded previous published HITS-CLIP lysis of XPO5-associated RNAs in human embryonic kidney cells (DROSHA and DGCR8)20,21, PAR-CLIP (DICER1) and AGO2/ 293T (HEK293T). We find that a large number of cellular RNAs 3-Seq (mature miRNAs)22 datasets and examined the profile of including most pre-miRNAs and numerous noncoding, structural miRNA sequences that are associated with each component RNAs are bound by XPO5. Notably, some closely clustered pri- (Fig. 1d–h). To facilitate the comparison of all miRNA profiles mary miRNA precursors such as pri-mir-17~92 and pri- captured by each CLIP dataset, we aligned all detected miRNA mir15b~16-2 show strong XPO5 HITS-CLIP signals outside of sequences starting from 5′ end of 5p miRNAs as annotated in pre-miRNA hairpins. Surprisingly, purified XPO5 directly bind to miRBase release 2123. Overall, DROSHA and DGCR8 HITS-CLIP pri-mir-17~92 and pri-mir15b~16-2 in a RanGTP-independent showed similar profiles, which usually span not only the pre- manner. In vitro processing assays reveal that XPO5 enhances the miRNA hairpin regions but also the flanking regions. In cleavage efficiency of pri-mir-19a and pri-mir-15b~16-2 by the agreement with the notion that DROSHA recognizes the lower DROSHA/DGCR8 microprocessor. stem from the basal junction that is outside of the pre-miRNA

2 NATURE COMMUNICATIONS | (2020) 11:1845 | https://doi.org/10.1038/s41467-020-15598-x | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15598-x ARTICLE

abc 1% 0% XPO5 IgG IP mir detected by smRNA-seq Loop Small RNA-seq Input + ++ +++ + RNase mir bound by XPO5 80 48% 70 50% 185 59 Left arm Left arm with loop 115 Right arm with loop 40 31 31 Right arm 80 1% Left arm Right arm 65 66 23.09% Complete 0 27.16% >10,000 Reads>1000 Reads>200 Reads XPO5 HITS-CLIP

50 hairpin detected pre-miRNA 7.41%

12.66%

miRNAs binned on abundance 29.68%

d DROSHA-CLIP efDGCR8-CLIP DICER-PARCLIP 0.8 0.8 0.8

0.6 0.6 0.6

0.4 0.4 0.4

0.2 0.2 0.2 Relative read density Relative read density Relative read density

0 0 0 –100 –50050 100 –100–50050 100 –100 –50 0 50 100 Distance relative to 5p miRNA +1 nucleotide Distance relative to 5p miRNA +1 nucleotide Distance relative to 5p miRNA +1 nucleotide

gh i AGO2/3-Seq XPO5-CLIP 0.8 0.8 XPO5-CLIP 0 (collapsed) –29 0 0.6 0.6 DROSHA-CLIP –1 0 0.4 0.4 DGCR8-CLIP –223 0 0.2 0.2 DICER-PARCLIP Relative read density Relative read density –16 AGO2/3-Seq 0 0 0 –100–50050 100 –100 –50 0 50 100 –606 Distance relative to 5p miRNA +1 nucleotide Distance relative to 5p miRNA +1 nucleotide MIR31HG mir31 j k 10,189 47 XPO5-CLIP XPO5-CLIP (all reads) 0 529 (collapsed) 0 XPO5-CLIP 42 (collapsed) 0 DROSHA-CLIP 451 0 DROSHA-CLIP 207 0 6097 DGCR8-CLIP DGCR8-CLIP 0 1 0 492 DICER-PARCLIP DICER-PARCLIP 0 505 0 6407 AGO2/3-Seq AGO2/3-Seq 0 0

DGCR8 Dgcr8

DGCR8 mir17 mir18a mir19a mir20a mir19b1mir92a1 mir3618 mir1306 hairpin and DGCR8 recognizes the upper stem of the miRNA pre-miRNA hairpin (Fig. 1f), consistent with that DICER1 hairpin24, DROSHA-associated miRNA sequences showed more recognizes pre-miRNA hairpin after its release by DROSHA/ prominent signals in the sequences immediately upstream of pre- DGCR8. AGO2/3-Seq showed the profile of mature miRNAs miRNA hairpin than DGCR8-associated miRNA sequences (Fig. 1g), which are derived from either the left or the right arm of (Fig. 1d, e). DICER1 PAR-CLIP showed the recognition of pre- miRNA hairpin. However, XPO5 HITS-CLIP showed not only miRNA hairpin but not the lower stem regions outside of the association of XPO5 with pre-miRNA hairpin as expected but

NATURE COMMUNICATIONS | (2020) 11:1845 | https://doi.org/10.1038/s41467-020-15598-x | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15598-x

Fig. 1 HITS-CLIP reveals association of XPO5 with primary and pre-miRNA precursors. a Autoradiogram of 32P-labelled XPO5-RNA complexes treated with different concentration of RNase is resolved on a PAGE gel. Original data are provided in the Source Data file. b Pre-miRNAs of most abundantly expressed miRNAs in HEK293T cells are detected in XPO5 HITS-CLIP. c The location of small RNA-seq reads and XPO5 HITS-CLIP reads relative to miRNA hairpin is shown in pie charts, respectively. Light blue indicates reads covering both the left arm and the loop region; blue indicates reads covering both the right arm and the loop region; dark blue indicates reads covering only the left arm; light gray indicates reads covering only the right arm; dark gray indicates reads covering the entire hairpin. d–h Metagene analysis of miRNA sequences associated by DROSHA, DGCR8, DICER1, AGO2/3, and XPO5 reveals the recognized regions by each core component and identifies pri-miRNA sequences associated by DROSHA, DGCR8, and XPO5 that are beyond pre-miRNA hairpin (red bracket). i, k IGV tracks of XPO5-CLIP, DROSHA-CLIP, DGCR8-CLIP, DICER1-PARCLIP, and AGO2/3-Seq reveal the specific association of XPO5 with pre-mir-31, pre-mir-3618, and pre-mir-1306. Negative reads number in IGV track indicates the mapping to the minus strand. j IGV tracks of XPO5-CLIP, DROSHA-CLIP, DGCR8-CLIP, DICER1-PARCLIP, and AGO2/3-Seq reveal the widespread association of XPO5 with pri-mir-17~92. Blue bars at the bottom and black lines together indicate the location of pre-miRNAs. Data range is shown on the left of each track.

also the flanking regions of the hairpin (Fig. 1h). Comparing lower level than pri-mir-17~92. We still observed prominent the overall profile of XPO5-associated miRNA sequences to inter-pre-miRNA reads in XPO5 HITS-CLIP data and the those of DICER1 and DGCR8, it was evident that XPO5- distinct reads density in the inter-pre-miRNA and pre-miRNA associated miRNA regions were similar to DICER1 within the regions (Supplementary Fig. 1d). Interestingly, the widespread annotated pre-miRNA hairpins but also resembled the profile of XPO5 reads coverage was not found in other miRNA clusters, in DGCR8-associated regions outside of the annotated pre-miRNAs which pre-miRNA hairpins are located far away from each other (Fig. 1e, f, h). such as the mir-200b cluster (>600 nt between hairpins) To understand how XPO5 associates with miRNA precursors (Supplementary Fig. 1e). These data suggest that not all primary outside of pre-miRNA hairpin, we next examined XPO5 HITS- transcripts of polycistronic miRNAs bind to XPO5. In addition, CLIP data of individual miRNAs. To faithfully present the data, two miRNAs, mir-3618 and mir-1306, are processed from the 5′ we used one IGV track to show all reads mapped from XPO5 region of Dgcr8 mRNA28. However, the expression of mir-1306 is HITS-CLIP data and another track to show all unique reads after much higher than mir-3618 based on both the processing assay28 collapsing potential duplicates (Fig. 1i–k and Supplementary and the sequencing of AGO2/3 associated mature miRNAs22. Fig. 1a, b). While many miRNAs are encoded by a primary Consistent with the pattern of mature miRNA expression, XPO5 transcript individually, ~50% of miRNAs are also encoded HITS-CLIP reads on mir-1306 pre-miRNA were much more by polycistronic transcription unit25,26. For all singly encoded abundant than those on mir-3618 pre-miRNA hairpin (Fig. 1k), miRNAs that we have examined such as mir-31, mir-21, and mir- lending further support to the specificity of our XPO5 HITS- 34a, XPO5 HITS-CLIP signals were faithfully restricted within CLIP. Together, these data suggest two modes of XPO5 pre-miRNA hairpins as expected (Fig. 1i and Supplementary association with miRNA precursors: (1) for most miRNAs, Fig. 1a, b). However, for many polycistronic miRNAs especially XPO5 associates with pre-miRNA hairpins as expected; and (2) the ones that are closely clustered, XPO5 HITS-CLIP reads also for some closely clustered polycistronic miRNAs, XPO5 associ- covered primary miRNA transcripts. Among all XPO5-associated ates with the primary miRNA transcripts with a high affinity in miRNAs, the mir-17~92 cluster harbored the most reads. inter-pre-miRNA regions. Remarkably, XPO5 HITS-CLIP reads broadly covered the pri- miRNA, distinct from the profile of pre-miRNA precursors as XPO5 binds to pri-miRNA precursors independently of detected in DICER1 PAR-CLIP (Fig. 1j). Inter-pre-miRNA RanGTP. To study the binding of XPO5 to closely clustered regions between mir-17 and mir-19b1 all had abundant XPO5- primary miRNAs, we performed electrophoretic mobility shift associated RNA reads compared with those derived from pre- assays in vitro. We purified human XPO5 and RanQ69L, a Ran miRNA hairpins (Fig. 1j). Upon closer inspection, we noted two mutant that is deficient in GTP hydrolysis and locks Ran in the possible modes of XPO5 binding to pri-mir-17~92, judging by the RanGTP form29, from bacteria E.coli as described previously4. distinct reads density covering the precursor regions. First, XPO5 We first used in vitro transcribed pre-mir-30a hairpin RNA to bound to pre-miRNA hairpins of each of these six miRNAs. confirm the binding activity of recombinant XPO5 and RanQ69L Second, XPO5 bound to inter-pre-miRNA regions with an even proteins to pre-miRNA. Indeed, XPO5 binds to pre-mir-30a higher density than the pre-miRNA regions. For example, both hairpin in a RanGTP-dependent manner (Fig. 2a). The single- mir-19a and mir-20a are abundantly expressed miRNAs. But shifted band also confirmed the formation of XPO5:RanGTP:pre- XPO5-associated RNA reads were far more abundant in the inter- mir-30a complex. Because pri-mir-17~92 is known to fold into a pre-miRNA regions than the pre-miRNA regions. Furthermore, highly structured RNA conformation that has extensive base- many XPO5-associated RNA reads span the predicted DROSHA pairing regions beyond pre-miRNA hairpins27 (Supplementary cleavage sites that mark the 5′ and 3′ ends of pre-miRNAs Fig. 2a), we in vitro transcribed and folded the pri-mir-17~92 (Supplementary Fig. 1c), indicating the association of XPO5 with precursor (Supplementary Fig. 2b) and tested its binding to uncleaved pri-miRNAs. The only inter-pre-miRNA region that is XPO5. Surprisingly, XPO5 bound to pri-mir-17~92 in a RanGTP- depleted of XPO5-associated RNA reads was between mir-19b1 independent manner (Fig. 2b). Inclusion of RanQ69L together and mir-92a1. Notably, mir-17~19b1 pri-miRNA is released from with XPO5 and pri-mir-17~92 showed no interference to the the mir-17~92 pri-miRNA by an endonuclease CPSF3 and the binding of XPO5 and the pri-miRNA (Fig. 2c). Neither the size spliceosome-associated ISY127. These data suggest a possibility nor the intensity of the super-shifted bands containing XPO5 and that XPO5 recognizes pri-mir-17~92 cluster after the CPSF3/ pri-mir-17~92 changed in the presence or absence of RanQ69L ISY1-mediated cleavage but prior to DROSHA/DGCR8-mediated (Fig. 2c). This result suggests that the complex does not contain processing. RanQ69L. In addition, we observed continuously super-shifted To confirm that the association of XPO5 to pri-miRNAs is bands with the increasing doses of XPO5, up to 80-fold of XPO5 specific and not simply due to the abundance of pri-mir17~92,we in excess of pri-mir-17~92 RNA (Fig. 2b, c). These data suggest inspected other closely clustered pri-miRNAs such as both mir- that multiple XPO5 molecules bind to pri-mir-17~92 precursor in 15a~16-1 and mir-15b~16-2 clusters that are expressed at a much a RanGTP-independent manner in vitro, corroborating with the

4 NATURE COMMUNICATIONS | (2020) 11:1845 | https://doi.org/10.1038/s41467-020-15598-x | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15598-x ARTICLE

abc XPO5 5×10×10×10× – XPO5 ––0.5×1× 2× 10× 40× 80×XPO5 1× 2× 10× 20× 40× 80× 20× RanQ69L – – +++ RanQ69L – ––––––RanQ69L ––++++++

Free RNA Free RNA Free RNA pre-mir-30a pri-mir-17~92a pri-mir-17~92a

d ef XPO5 5× 10× 20×15× RanQ69L – – ++ XPO5 – 1× 5× 10× 20× 40× 60× 80×100×150×200× – pri-mir-19a 100,000

50,000 Kd = 14.56 nM Kd = 14.56 nM Free

RNA Signal strength 0 0 10 20 30 40 50 XPO5 (nM) Free RNA pri-mir-19a pre-mir-19a

g hi XPO5 – 1× 5× 10× 40× 60× 80× 100× XPO5 ––1× 2.5× 5× 10×20×40×60×80×100×150×200×

pri-mir-19a truncation v1 200,000

Free 100,000 RNA Kd = 7.99 nM free Signal strength 0 RNA 050100150

pri-mir-19a truncation v1 XPO5 (nM) pri-mir-19a truncation v2

Fig. 2 XPO5 binds to primary miRNA precursors in a RanGTP-independent manner. a XPO5 binds to pre-mir-30a in a RanGTP-dependent manner. Black hairpin represents pre-mir-30a, XPO5 is coloured in orange, RanGTP is coloured in purple. b XPO5 binds to pri-mir-17~92 without RanGTP. Increasing amount of XPO5 results in super shifts. c XPO5 binding to pri-mir-17~92 is not affected by RanGTP. d XPO5 binds to pre-mir-19a in a RanGTP-dependent manner. Black hairpin represents pre-mir-19a, XPO5 is coloured in orange, RanGTP is coloured in purple. e XPO5 binds to pri-mir-19a in a RanGTP- independent manner. Increasing amount of XPO5 results in a super shift. Black hairpin represents pri-mir-19a, XPO5 is coloured in orange. f The dissociation constant (Kd) between XPO5 and pri-miR-19a is calculated based on the binding results in e. g XPO5 binds to truncated pri-mir-19a v1 (pri-mir- 19a truncation v1) in a RanGTP-independent manner. Black hairpin represents pri-mir-19a truncation v1, XPO5 is coloured in orange. h The dissociation constant (Kd) between XPO5 and pri-miR-19a truncation v1 is calculated based on the binding results in g. i XPO5 does not bind to truncated pri-mir-19a v2 (pri-mir-19a truncation v2). For all binding assays, 1 nM substrates are used. Original data for a–e, g, and i are provided in the Source Data file. extensive XPO5 HITS-CLIP signals detected on pri-mir-17~92 lower stem region that is characteristic of DROSHA/DGCR8 in vivo. primary miRNA substrates7 (Supplementary Fig. 2c). The To further characterize how XPO5 binds to primary miRNA primary miRNA sequences flanking mir-19a also had very high transcript, we focused on pri-mir-19a, whose flanking sequences XPO5 HITS-CLIP signals (Fig. 1j). We first confirmed that pre- are predicted to form extensive base-pairing outside of the 11 bp mir-19a hairpin only bound to XPO5 in a RanGTP-dependent

NATURE COMMUNICATIONS | (2020) 11:1845 | https://doi.org/10.1038/s41467-020-15598-x | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15598-x manner (Fig. 2d). In contrast, pri-mir-19a bound to XPO5 in a XPO5, was not further enhanced by XPO5 (Fig. 3c). Together, RanGTP-independent manner, similar to the pri-mir-17~92 these data reveal that RanGTP-independent binding of pri-mir- results. Importantly, two distinct pri-mir-19a:XPO5 complexes, 19a by XPO5 outside of the pre-miRNA hairpin promotes the judging by the size of super-shifted pri-mir-19a, were detected microprocessor cleavage in intro. with the increasing amount of XPO5 (Fig. 2e). Finally, the dissociation constant (Kd) of pri-mir-19a and XPO5 association XPO5 promotes pri-miRNA processing of clustered miRNAs. was calculated at 14.56 nM (Fig. 2f). To extend these findings beyond pri-mir-19a, we turned to pri- Because pre-mir-19a hairpin could not bind to XPO5 in the mir-15b~16-2, which encodes two miRNAs and also has extensive absence of RanGTP, these results suggested that XPO5 binds to XPO5 association outside of pre-miRNA hairpin (Supplementary the lower stem regions of pri-mir-19a. To test this hypothesis, we Fig. 1d). Notably, the predicted secondary structure of pri-mir- generated two truncated pri-mir-19a mutants. We deleted one 15b~16-2 also contains extensive base-pairing regions outside of basal stem region in pri-mir-19a truncation v1 and both basal the pre-miRNA hairpin (Fig. 4a). In support of the HITS-CLIP stem regions in pri-mir-19a truncation v2 (Supplementary data, pri-mir-15b~16-2 bound to XPO5 in a RanGTP- Fig. 2c). Indeed, pri-mir-19a truncation v1 bound to XPO5 in a independent manner. Super-shifted bands were also observed RanGTP-independent manner with the Kd of 7.99 nM but only a with the increasing dose of XPO5 (Fig. 4b), similar to the pattern singly shifted band was detected (Fig. 2g, h). In contrast, pri-mir- of pri-mir-19a. In addition, an isolated pri-mir-15b containing the 19a truncation v2 no longer bound to XPO5 (Fig. 2i) when both lower stem region also bound to XPO5 and formed a singly basal stem regions were deleted. These data suggest that the shifted complex with the Kd of 9.57 nM (Fig. 4c–e), similar to the RanGTP-independent binding of pri-mir-19a by XPO5 requires pattern of pri-mir-19a truncation v1 (Fig. 2g, h). Together, these the lower stem regions outside of the pre-miRNA hairpin. in vitro data reveal an unexpected RanGTP-independent binding of XPO5 to pri-mir-17~92 and pri-mir-15b~16-2 polycistronic miRNA, pri-mir-19a and pri-mir-15b precursors. XPO5 facilitates the microprocessor cleavage of pri-miRNAs. Because of the RanGTP-independent association of pri-mir- To determine the function of XPO5 binding to pri-mir-19a pre- 15b~16-2 and pri-mir-15b to XPO5 (Figs. 4b, d), we next tested cursors, we performed the DROSHA/DGCR8-mediated proces- these primary miRNAs with the processing assay. The processing sing assay for pri-mir-19a and the two mutants in vitro. Flag- efficiency of both pri-mir-15b~16-2 and pri-mir-15b by the tagged DROSHA and DGCR8 expression plasmids were microprocessor was also strongly enhanced (approximately co-transfected into HEK293T cells to obtain the microprocessor twofold) by the presence of XPO5 (Fig. 4f, lanes 3–6 and Fig. 4g, complex using immunoprecipitation and purification as described lanes 3–5). Notably, the unprocessed pri-mir-15b~16-2 and pri- previoulsy30. We also purified Flag-XPO5 protein from trans- mir-15b was strongly reduced. We detected both singly cropped fected HEK293T cells. Western blot and silver stain indicated that pri-mir-15b~16-2 intermediates and pre-mir-15b/16-2, and the DROSHA/DGCR8 microprocessor does not copurify with XPO5 accumulation of both was increased by XPO5. In addition, the (Supplementary Fig. 3a, b). In the absence of XPO5, the micro- quadruple XPO5 mutant also promoted the DROSHA/DGCR8 processor complex released pre-mir-19a hairpin from the primary processing (Fig. 4f, lanes 8–11). Taken together, these data transcript with a modest efficiency (Fig. 3a, lane 3), in line with suggest that RanGTP-independent binding of pri-mir-19a, pri- previously studies30,31. Notably, preincubation of pri-mir-19a mir-15b~16-2 and pri-mir-15b by XPO5 facilitates the DROSHA/ with an increasing amount of XPO5 drastically enhanced the DGCR8 processing. These findings reveal an additional function processing efficiency. The improved processing was demonstrated of XPO5 independent of nuclear export. by both the significant reduction of unprocessed pri-mir-19a and approximately twofold more accumulation of pre-mir-19a in an XPO5 dosage-dependent manner (Fig. 3a, lanes 4–6). XPO5 alone XPO5 binds to cellular RNAs with double-stranded regions.In did not alter pri-mir-19a (Fig. 3a, lane 7), ruling out the possi- addition to miRNA precursors, we identified numerous non- bility of microprocessor contamination. coding RNAs in the XPO5 HITS-CLIP datasets (Fig. 5a and XPO5 has been well characterized for its binding to pre-miRNA Supplementary Data 2). Among these, noncoding RNAs were hairpin that is dependent upon RanGTP and 2-nt 3′ overhang of SINE and LINE elements (repeat), vault RNAs (vtRNAs), 7SL pre-miRNA6,32. So far, we have showed the RanGTP-independent RNA, snRNA, snoRNA, 7SK RNA, tRNA, and hTR (Fig. 5b, g, i binding of pri-miRNA by XPO5 (Fig. 2b–e, g). We next asked and Supplementary Fig. 4). vtRNAs were among the most whether XPO5 residues that mediate the interaction between abundant XPO5-associated RNAs. Notably, vtRNAs were abun- XPO5 and the 5′ and 3′ ends of pre-miRNA are required for dantly bound by all core components of miRNA biogenesis and enhancing the microprocessor cleavage. To this end, we generated likely gave rise to small RNAs associated with AGO proteins an XPO5 mutant with E445A, S606A, E711A, and R718A (Fig. 5b). Similar to pri-mir-19a, vtRNAs were bound by XPO5 in quadruple mutations. These residues were found to form a RanGTP-independent manner (Fig. 5c). Furthermore, although hydrogen bonds between XPO5 HEAT repeats and the 5′ and vtRNAs typically formed one major complex with XPO5, vtRNAs 3′ ends of pre-mir-30a including the 2-nt 3′ overhang6. In support were also super-shifted by the increasing amount of XPO5 pro- of the notion that XPO5 interaction with pri-miRNA is tein (Fig. 5c). The Kd of vtRNA and XPO5 association was independent of its association with pre-miRNA, the XPO5 determined at 36.27 nM (Fig. 4d), weaker than that of pri-mir-19a mutant retained the ability to potentiate the DROSHA/DGCR8 and pri-mir-15b. The weaker interaction between vtRNAs and processing (Fig. 3a, lanes 8–11). XPO5 was correlated with the less prominent and shorter dsRNA To further confirm that the ability of XPO5 to promote pri- structures of vtRNAs (Fig. 5e and Supplementary Fig. 5a). mir-19a processing requires RanGTP-independent binding, we To probe the potential effect of XPO5 on the biogenesis and performed the same assays using the two truncated mutants. localization of vtRNA, we generated XPO5 knockout (KO) MEF Importantly, although both mutants can be cleaved by the cells (Supplementary Fig. 5b, c, also see below). Interestingly, microprocessor, only pri-mir-19a truncation v1, which is bound neither the level nor the cytoplasmic localization pattern of by XPO5 in the absence of RanGTP, was more efficiently cleaved vtRNA was altered in the absence of XPO5 (Fig. 5f). These data by both the WT and mutant XPO5 (Fig. 3b). In contrast, the suggest that the binding of vtRNA by XPO5 does not affect the processing of pri-mir-19a truncation v2, which is not bound by maturation, accumulation, and localization of vtRNA.

6 NATURE COMMUNICATIONS | (2020) 11:1845 | https://doi.org/10.1038/s41467-020-15598-x | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15598-x ARTICLE

a Lane 1 2 3 4 5 6 7 8 9 10 11 12 c + + + + + + + + DROSHA/DGCR8 Lane 1 2 3 4 5 6 0.5× 1× 2× 1× XPO5 + + + DROSHA/DGCR8 1× 2× 1× XPO5 M 0.5× 1× 2× 1× XPO5 mutant M 1 1 0.8 0.4 0.3 0.2 1 1 0.5 0.2 0.2 1 (Quan.) (nt) 1 0.9 0.8 0.7 0.9 0.9 (Quan.) (nt) 150- pri-mir-19a 150-

100- pri-mir-19a 100- truncation v2

pre-mir-19a

pre-mir-19a

0 0 1 0.7 1.1 (Quan.) 0 0 1 1.9 2.2 2.3 0 1 1.7 1.1 1.4 0 (Quan.)

b Lane 1 2 3 4 5 6 7 8 9 10 11 12 13 14 + + + + + + + + + + DROSHA/DGCR8 0.5× 1× 2× 1× 1× XPO5 M 0.5× 1× 2× 1× 1× XPO5 mutant

(nt) 1 1 0.7 0.1 0.1 0.1 0.1 0.9 0.6 0.1 0.1 0.1 0.1 0.6 (Quan.)

150- pri-mir-19a trucation v1

100-

pre-mir-19a

0 0 1 2.2 2.6 2.4 2.5 0 1 2 2.3 2.6 2.4 0 (Quan.)

Fig. 3 XPO5 facilitates the microprocessor cleavage of pri-mir-19a.aPreincubation of XPO5 enhances the microprocessor cleavage of pri-mir-19a. Incubation of increasing amount of XPO5 leads to increased processing efficiency of pri-mir-19a (lanes 3–6). Incubation of increasing amount of the XPO5 mutant also results in increased processing efficiency of pri-mir-19a (lanes 8–11). b Preincubation of XPO5 and XPO5 mutant enhances the microprocessor cleavage of truncated pri-mir-19a v1. c Preincubation of increasing amounts of XPO5 does not promote the cleavage efficiency of pri-mir-19a truncation v2 by the microprocessor. Representative images of three independent repeats are shown for each assay (a–c). (Quan. = quantification). Original data for all panels are provided in the Source Data file. .

We noticed that many of XPO5-associated RNAs are predicted scaRNA domain, respectively (Fig. 5g). In addition, AGO to form double-stranded regions. In general, XPO5 showed more interacts with small RNA fragments that are derived from widespread binding patterns than the other core components the scaRNA domain located at the 3′ end of hTR but not the more including DROSHA, DGCR8, and DICER1 in structured RNAs prominent template region or the CR4/5 domain (Fig. 5g). These such as hTR and 7SL RNA (Fig. 5g–i). hTR interacts with XPO5 data suggest that XPO5 preferentially interacts with structured in three domains with the strongest interaction detected towards regions of hTR. Similar to hTR, 7SL RNA has well-defined the 5′ end of hTR. Intriguingly, the number of XPO5 HITS-CLIP secondary structures with extensive base-pairing regions that are reads associated with each domain was different, despite of their conserved through evolution35,36. Interestingly, 7SL RNA origin from the same RNA (Fig. 5g). To probe whether XPO5 strongly interacts only with XPO5 but not with other core association may be correlated with the strength of base-pairing, components and AGO-associated small RNA fragments were not we further examined XPO5-bound regions on hTR. Three highly detected (Fig. 5i), indicating a possible function of XPO5 that is structured regions of hTR including the template region, the CR4/ independent of miRNA biogenesis. CR5 domain and the scaRNA domain have been determined To globally measure the preference of XPO5 to base-paired previously33,34. Among these regions, the template region forms RNAs, we performed an RNA Folding energy analysis for all the most extensive dsRNA structures, followed by the CR4/CR5 XPO5-associated RNAs based on HITS-CLIP. We classified these and scaRNA domains (Fig. 5h). Notably, XPO5-associated RNA RNAs into repeat associated regions, nonrepeat associated regions reads were mostly enriched in the double-stranded regions (P2a- and pre-miRNAs. As shown in Fig. 5j, the folding energy of P2b-J2b/3) of the template region, followed by the reads over the XPO5-assoicated pre-miRNAs was significantly lower than that double-stranded regions of the CR4/CR5 domain and the of randomly selected sequences and indistinguishable from that

NATURE COMMUNICATIONS | (2020) 11:1845 | https://doi.org/10.1038/s41467-020-15598-x | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15598-x

a b XPO5 0.1× 0.5× 1× 5× 20×

Secondary structure of pri-mir-15b~16-2 by RNAfold

pre-mir-15b

pre-mir-16-2

or 01 Free RNA

pri-mir-15b~16-2

c d XPO5 __ 2.5× 5×10× 20× 40× 60× 80×100×150× e

Secondary structure of pri-mir-15b by RNAfold pri-mir-15b pre-mir-15b 200,000

100,000 Kd = 9.57 nM

01 Signal strength 0 Lower stem 0 50 100 150 200 Free RNA XPO5 (nM) > 11 bp

pri-mir-15b

f Lane 1 2 3 4 5 6 7 8 9 10 11 12 g + + + + + + + + DROSHA/DGCR8 Lane 1 2 3 4 5 6 0.5× 1× 2× 1× XPO5 + + + DROSHA/DGCR8 XPO5 0.5× 1× 2× 1× XPO5 mutant 1× 2× 1× M M 1 0.9 0.7 0.2 0.3 0.2 1.3 1 0.6 0.3 0.3 1.1 (Quan.) 1 1 0.8 0.3 0.3 1 (Quan.) pri-mir-15b-16-2 (nt)

Singly cropped pri-mir-15b pri-mir-15b-16-2 100- 0 0 1 4 4.5 4.8 0 1 4.8 5.1 5.4 0 (Quan.) (nt)

60- pre-mir-15b 100-

pre-mir-15b/16-2 0 0 1 1.4 1.4 0 (Quan.) 60-

0 0 1 1.8 2.2 2.4 0 1.2 2.1 2.3 2.4 0 (Quan.)

Fig. 4 XPO5 promotes primary miRNA processing of clustered miRNAs. a Secondary structure of pri-mir-15b~16-2 is predicted by RNAfold. Heat map indicates the probability of predicted structures (purple: 0—low; red: 1—high). b XPO5 binds to pri-mir-15b~16-2 in a RanGTP-independent manner. Increasing amount of XPO5 results in a super shift. Black hairpin represents pri-mir-15b~16-2. XPO5 is coloured in orange. c Secondary structure of pri-mir- 15b is predicted by RNAfold. d XPO5 binds to pri-mir-15b in a RanGTP-independent manner. Black hairpin represents pri-mir-15b. XPO5 is coloured in orange. e The dissociation constant (Kd) between XPO5 and pri-miR-15b is calculated based on the binding results in d. f Preincubation of XPO5 enhances the microprocessor cleavage of pri-mir-15b~16-2. Incubation of increasing amount of XPO5 leads to increased processing efficiency of pri-mir-15b~16-2 (lanes 3–6). Incubation of increasing amount of the XPO5 mutant also results in increased processing efficiency of pri-mir-15b~16-2 (lanes 8–11). g Preincubation of increasing amounts of XPO5 promotes the cleavage efficiency of pri-mir-15b by the microprocessor. Representative images of three independent repeats are shown for each assay (f–g). (Quan. = quantification). Original data for b, d, f, and g are provided in the Source Data file.

8 NATURE COMMUNICATIONS | (2020) 11:1845 | https://doi.org/10.1038/s41467-020-15598-x | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15598-x ARTICLE

a b 100 chr5:140,090,857–140,090,965 26,945 XPO5-CLIP (all reads) 0 XPO5-CLIP 758 75 Intron (1270) (collasped) 0 Repeat (794) 23 ncRNA (374) DROSHA-CLIP 0 tRNA (237) 71 50 3utr (202) DGCR8-CLIP 0 None (199) 1068 cds (126) DICER-PARCLIP with annotation 0 Percent of peaks Other (116) 336 25 AGO2/3-Seq pre_miR (107) 0 snoRNA (87) vtRNA1-1

0 Xpo5Peaks 2: CR4/5 domain c __ g chr3:169,482,358–169,482,906 XPO5 5×10×15×20×40×80×100×150× Domains: 3 2 1

0 XPO5-CLIP (all reads) -403 XPO5-CLIP 0 (collasped) -93 DROSHA-CLIP 0

DGCR8-CLIP 0

DICER-PARCLIP 0 -1 Free AGO2/3-Seq 0 vtRNA1-1 -33 hTR d e h vtRNA1-1 Predicted 60,000 secondary 1: Template region 40,000 structure of vtRNA1-1 20,000 Kd = 36.27 nM 3: ScaRNA domain Signal strength 0 0205080100 Template region XPO5 (nM) 01 p = 2.51e – 109 p = 1.67e – 14 f i j p = 1.82e – 49 fl/fl p = 1.06e – 9 p = 0.96 p = 2.39e – 48 chr14:50,053,297–50,053,649 17,864 0.00 XPO5 XPO5KO-1XPO5KO-2 XPO5-CLIP (all reads) 0 -0.25 MEF NCNC NC XPO5-CLIP 740 0 -0.50 (collasped) 11 vtRNA DROSHA-CLIP 0 -0.75

DGCR8-CLIP per nucleotides -0.10

0 G 0.1 1 0.1 1.1 0.1 1.2 6 Δ DICER-PARCLIP 0 5S RNA AGO2/3-Seq 0 XPO5 peaksXPO5 peaksXPO5 peaks ffled control (repeat ) (non-repeat)pre-miRNA 0.2 1 0.3 0.8 0.3 0.7 Shu 7SL RNA1 miRbase pre-miRNA of all pre-miRNA hairpins. Interestingly, both repeated associated the function of XPO5 in mouse development. We identified a and nonrepeat associated regions bound by XPO5 had an even knock-in (KI) first, conditionally targeted embryonic stem lower folding energy than that of pre-miRNA hairpins (Fig. 5j). cell clone for generating XPO5 KO mouse models37. Founder KI These data suggest that XPO5 has a preference to double- mice were produced and validated for germline transmission of stranded regions of cellular RNAs. the targeted allele that contains an IRES-LacZ and Neo cassette in the intron 7 and a floxed exon 8 of XPO5 (Fig. 6a). To generate the conditional allele, the founder KI mice were bred to Actb-Flp XPO5 is required for mouse embryonic development. Having mice to remove the IRES-LacZ and Neo cassette. Finally, the characterized XPO5-associated cellular RNAs, we next examined conditional allele was bred to EIIA-Cre or Krt14-Cre mice to

NATURE COMMUNICATIONS | (2020) 11:1845 | https://doi.org/10.1038/s41467-020-15598-x | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15598-x

Fig. 5 XPO5 associates with diverse cellular RNAs. a Peak annotation of XPO5-associated reads from XPO5 HITS-CLIP. b, g, i IGV tracks of vtRNA1-1, hTR, and 7SL RNA1 show the association of XPO5 with each noncoding RNA, compared with DROSHA, DGCR8, DICER1, and AGO2/3. Blue bar at the bottom indicates the coding region and the arrow indicates the direction of transcription. Data range is shown on the right of each track. c XPO5 binds to vtRNA1-1 without RanGTP. 1 nM vtRNA1-1 substrate was used. Red arrows indicate vtRNA1-1 RNA with (upper) and without (lower) XPO5 binding, red bracket indicates the super-shifted vtRNA1-1 RNA with more than one XPO5 molecule. d The dissociation constant (Kd) between XPO5 and vtRNA1-1 is calculated based on the binding results in c. e Predicted secondary structure of vtRNA1-1. The potential to form predicated structures is coloured (purple to red → low to high). f Detection of vtRNA and 5S RNA (loading control) from nuclear and cytoplasmic RNAs extracted from wild type and XPO5 KO MEF cells by northern blotting. h XPO5 HITS-CLIP reads align to the secondary structure of hTR. Red and green arrows indicate the 5′ and 3′ location of XPO5- associated RNA regions, respectively. Black lines outline the XPO5-associated hTR regions. j An RNA Folding analysis shows the folding energy of XPO5- assoicated repeat RNAs, XPO5-assocatied nonrepeat RNAs, XPO5-associated pre-miRNAs, miRbase pre-miRNAs, and randomly shuttled control sequences. For each boxplot, the middle line is the median, the vertical line spans the data range, and the hinges are the first and third quartiles. A two- sided Mann–Whitney test was used for statistical tests. Original data for c and f are provided in the Source Data file.

generate constitutive KO and skin cKO of XPO5, respectively terminally differentiated inner root sheath and hair shaft, LacZ (Fig. 6a). signals appeared to be elevated in these differentiated cells than Neither XPO5KI/+ nor XPO5KO/+ breeding produced any their progenitor counterparts (Fig. 7b). In XPO5 skin cKO viable XPO5 null (XPO5KI/KI or XPO5KO/KO) mice at birth, animals, we observed progressively compromised hair follicle whereas XPO5 WT and heterozygous mice were obtained at the development that was similar but slightly milder, compared with expected Mendelian ratio, indicating a requirement of XPO5 Dicer1 and Dgcr8 cKO (Fig. 7c). XPO5 cKO animals showed during mouse embryonic development (Fig. 6b). Closer inspec- neonatal lethality between postnatal day 5 and 18 and failed to tion revealed that XPO5 null embryos showed signs of gain weight postnatally (Fig. 7c–e). This was in contrast to Ago1/2 compromised development around embryonic day 6.5, a devel- dKO animals, which lose ~80% of total miRNAs and usually opmental stage that is correlated with the onset of gastrulation survive to adulthood despite the apparent loss of hair follicles12. and embryonic germline layer formation. At this stage, XPO5 null Consistent with these observations, qRT-PCR quantification of embryos were much smaller, compared with their control mir-203 and mir-205, two of the most highly expressed miRNAs littermates (Fig. 6c). Furthermore, XPO5 null embryos failed to in the skin, showed >90% depletion for both miRNAs (Fig. 7f). In go through gastrulation and initiate organogenesis by E8.5 addition, AGO2 protein was also depleted in XPO5 null skin (Fig. 6d). Judging by the LacZ signals in E7 XPO5KI/+ and (Fig. 7g). Because mature miRNA accumulation is required to XPO5KI/KI embryos, XPO5 was universally expressed in all stabilize AGO proteins40, this result provided further support to embryonic cells (Fig. 6e). the strong depletion of global miRNA expression in the absence We then performed quantitative real-time PCR (qRT-PCR) of XPO5. analyses to validate the loss of XPO5 and mir-290, a representa- To further document the role of XPO5 during skin morpho- tive miRNA that is highly expressed in mouse ESCs and early genesis, we examined neonatal skin. We observed reduced hair embryos, using total RNA isolated from E7 WT, het, and KO follicle formation and stunted hair follicle down growth in embryos (Fig. 6f). Because the miRNA pathway is required to P2 skin (Fig. 7h). When examined specifically with Lef1 and β4- promote the differentiation of ESCs to lineage specific cell types integrin staining for hair germ, we found many evaginating hair and the loss of Dicer1 and Dgcr8 blocks ESC differentiation38,39, germs towards the epidermis that is characteristic of Dicer1 and we also quantified ESC marker genes such as Nanog and Oct4 as Dgcr8 cKO skin17–19 (Fig. 7i). Furthermore, epidermal cell well as differentiation markers such as T and Tet1 by qRT-PCR. proliferation was strongly reduced (Fig. 7j). Finally, epidermal The results revealed a strong accumulation of Nanog and Oct4 differentiation was not strongly affected as documented by the mRNAs in the E7 XPO5 null embryos, and a relatively normal Krt5 and Krt1 staining that illuminated the basal and differ- level of T and Tet1 (Fig. 6g). Furthermore, the high level of Oct4 entiated spinous layer, respectively (Fig. 7k) and by qRT-PCR persisted in E8.5 XPO5 null embryos (Fig. 6h), reflecting a analysis (Fig. 7l). Collectively, these studies provide further compromised differentiation. Together, these data reveal that genetic evidence to support the requirement of XPO5 for miRNA XPO5 is required for miRNA biogenesis and mouse embryonic biogenesis and skin development in mouse. development. XPO5 is required for miRNA biogenesis. Having demonstrated XPO5 is required for skin morphogenesis. We have previously the requirement of XPO5 in mouse embryonic and tissue studied Dicer1, Dgcr8, as well as Ago1/2 in the skin using a K14- development, we further characterized the global miRNA Cre line and observed unique defects in hair morphogenesis that expression levels in XPO5 KO skin using a quantitative small are characteristic of the loss of the entire miRNA pathway in the RNA sequencing method41 (Fig. 8a). Globally, small RNAs skin12,18,19. Furthermore, both Dicer1 and Dgcr8 cKO skin typi- between 20 and 23 nt showed a strong depletion in the XPO5 cally loses >95% of individual miRNAs with a few exceptions of KO samples (Fig. 8b). We then specifically examined miRNA short hairpin miRNAs such as mir-320 and mir-484 that are depletion. Although it was somewhat weaker than the deple- dependent on Dicer1 but independent of Dgcr819. In comparison, tionobservedintheDicer1 and Dgcr8 cKO skin (Fig. 8c), the Ago1/2 dKO skin loses ~80% of individual miRNAs and shows average level of depletion was ~90%. The depletion of mature milder defects12. To compare the developmental defects of XPO5 miRNA was particularly evident for highly expressed miRNAs cKO animals with these well-characterized cKO models of the whose expression can be more robustly measured by small miRNA biogenesis pathway in the skin, we generated K14-Cre/ RNA sequencing (Fig. 8d and Supplementary Data 3). In XPO5fl/fl mice (Fig. 7a). addition to highly expressed mir-203 and mir-205 (Fig. 7f), we We first documented the expression pattern of XPO5 in the quantified the depletion of five miRNAs from the mir-17~92 skin by monitoring LacZ signals in XPO5KI/+ heterozygous mice. and mir-15~16 clusters (Fig. 8e, f), which expressed at an XPO5 was universally expressed in all skin cell types at postnatal intermediate level in the skin. We observed strong depletion day 2 and 4 (P2 and P4). By P4, when hair follicles gave rise to for mir-17-5p, mir-18, mir-19a/b,andmir-20a except mir-92a

10 NATURE COMMUNICATIONS | (2020) 11:1845 | https://doi.org/10.1038/s41467-020-15598-x | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15598-x ARTICLE

ab KO/+ breeding KI/+ breeding KI IRES- neo 8 9 XPO5 7 LacZ (20 litters) (36 litters) +Flipase GT # GT # XPO5fl Frt 7 8 9 WT 33 WT 51 LoxP +Cre KO/+ 80 KI/+ 96 cKO XPO5 79 KO/KO 0 KI/KI 0

c E6.5 d E8.5

Ctrl KO Ctrl KO ef E7 WT Het KO 2

1.5

Cq) 11 1 ΔΔ

0.5 *** *** .14

Relative RNA level ( .02 0 XPO5 miR290 Ctrl KO gh E7 WT Het KO E8.5 Ctrl KO 20 ** 14 ** 12 15 10 Cq) Cq) 8 ΔΔ ΔΔ

( 10 6 5 * 4 Relative RNA level 1 Relative RNA level ( 2 *** .02 0 0 Nanog Tet1 T Oct4 XPO5 Oct4

Fig. 6 XPO5 is required for mouse embryonic development. a Schematics of the XPO5 allele design. b Loss of XPO5 results in embryonic lethality. XPO5 KO embryos show developmental defects as early as in E6.5 (c) and fails to go through gastrulation (d). Representative images of littermate embryos are shown for over ten control and KO embryos. e Whole-mount LacZ staining of XPO5KI/+ (Ctrl) and XPO5KI/KI (KO) shows universal expression of XPO5 throughout the embryo. f Depletion of XPO5 and mir-290 in XPO5 KO embryos at E7 is measured by qPCR. Dysregulation of selected genes in XPO5 KO embryos at E7 (g) and E8.5 (h) is measured by qPCR, respectively. Oct4 and Nanog are failed to be downregulated in XPO5 KO embryos at both E7 and E8.5. Scale bar 250 μminc–e. Data shown are mean s.d. from three independent experiments. *P < 0.05; **P < 0.01; ***P < 0.001 determined by Student’s t test. Original data for f, g, and h are provided in the Source Data file. as well as mir-15a/b and mir-16 (Fig. 8e, f). Interestingly, the independent of Drosha/Dgcr8 processing19,werelargely loss of mir-19a/b was particularly strong, likely reflecting the unchanged in XPO5 cKO skin (Fig. 8h), lending support to the requirement of XPO5 for its processing and nuclear export. To notion that XPO5 is required to export pre-miRNA hairpin further validate the compromised miRNA biogenesis and dis- generated by the Drosha/Dgcr8 microprocessor. Taken toge- tinguish the effect of XPO5 loss on different miRNA species, ther, these analyses reveal the requirement of XPO5 for the we performed northern blotting for mir-17 (Fig. 8g). We biogenesis of a majority of miRNAs in the skin, consistent with observed the loss of mature mir-17 but no strong changes of the strong defects observed in the cKO model. We also note, pre-mir-17 were detected (Fig. 8g). The lack of accumulation of however, whether the phenotypes of XPO5 KO is entirely due pre-miRNAs is likely due to their unstable nature in the to the loss of miRNAs should be further examined due to the absence of XPO5, similar to the patterns reported in Dicer1 binding of XPO5 to many cellular RNAs with double-stranded cKO18. In addition, mir-320 and mir-484, whose biogenesis is regions.

NATURE COMMUNICATIONS | (2020) 11:1845 | https://doi.org/10.1038/s41467-020-15598-x | www.nature.com/naturecommunications 11 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15598-x

a c P1.5 P4 P11 P14 Ctrl cKO Ctrl cKO Ctrl Ctrl XPO5fl 789

+Krt14-Cre

XPO5cKO 79 cKO cKO

b de KI 100 XPO5 14 80 12 Ctrl (n = 3) P2 10 cKO (n = 5) 60 8 40 6

Ctrl (n = 3) Mass (g) 4

Percent survival 20 cKOs (n = 5) 2 0 0 6 7 8 10 11 12 13 14 15 18 P4 0 5 10 15 20 Postnatal day Postnatal day fg Ctrl cKO fl/fl cKO

1.5 XPO5 XPO5

Cq) 1 XPO5 ΔΔ 0.5 Tubulin *** *** ***

Relative RNA level ( 0 AGO2 XPO5 mir-203 mir-205 Tubulin h i Ctrl cKO 4/Lef1 Ctrl

cKO

j k l Ctrl cKO NS Ctrl Ctrl 1.2 NS NS 1.0 Cq) * 0.8

Krt 1 ΔΔ  4/Ki67 Krt 5 0.6 0.4 0.2 cKO cKO Relative RNA level ( 0 Krt1 Lor Krt5 Itgb4

Fig. 7 XPO5 is required for skin development. a Schematic of K14-Cre/XPO5fl/fl mouse model. b Whole-mount LacZ staining of XPO5KI/+ in P2 and P4 skin. c Neonatal mice of XPO5 cKO animals with littermate controls from P1.5 to P14. d, e XPO5 cKO animals are neonatal lethal and fail to gain weight. f Depletion of mir-203 and mir-205 in XPO5 cKO epidermis is quantified by qPCR. Data shown are mean s.d. from three independent experiments. ***P < 0.001 by Student’s t test. g Depletion of XPO5 and AGO2 proteins in XPO5 cKO epidermis is confirmed by western blot. h H&E staining reveals reduced hair follicle formation and stunted hair follicle growth in XPO5 cKO skin at P2. i Stunted hair follicle down growth and evaginating hair germ (arrowhead) in XPO5 cKO skin are revealed by Lef1 and β4-integrin staining. j Reduced cell proliferation in XPO5 cKO skin is shown by Ki67 staining. k XPO5 cKO skin shows normal epidermal differentiation as determined by Krt5 and Krt1 staining. l Normal epidermal differentiation is confirmed by qPCR detection of Krt1, Lor, Krt5, and Itgb4 (encodes β4-integrin). Data shown are mean s.d. from three independent experiments. *P < 0.05, n.s.—not significant, by Student’s t test. Scale bar: 1 cm in c, 100 μminb, h, j, and k,50μmini. Original data for d–g and l are provided in the Source Data file.

Discussion miRNAs (Fig. 1b). Surprisingly, however, XPO5 strongly binds In this study, we have determined XPO5-associated RNA spe- to primary miRNA precursors of some closely clustered poly- cies at the genomic scale. As expected, XPO5 associates with cistronic miRNAs such as mir-17~92 and mir-15b~16-2 pre-miRNA precursors for a vast majority of expressed (Fig. 1i–k and Supplementary Fig. 1). The in vitro binding

12 NATURE COMMUNICATIONS | (2020) 11:1845 | https://doi.org/10.1038/s41467-020-15598-x | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15598-x ARTICLE

abSmall RNA-Seq on RNA from Total epidermis (p2–p4) with synthetic RNA spike in controls 0.6 0.5 Cont.

Align Bowtie2 0.4 cKO 0.3

mapped) 0.2 Count reads overlap

with miRNA (miRbase v21), Reads proportion (mapped miRs/All 0.1 snRNA, snoRNA, tRNA, smRNA, amd rRNA (Gencode M8) annotations 0 18 19 20 21 22 23 24 25 26 27 Mapped read length(nt)

Compute Log2 fold change c d miRNA Non-miRNA miRNA Non-miRNA

5e – 15 3e – 14 5e – 69 2 5 10

5 0 0

–5 –5 mean expression (Log) mean expression Log2 fold change Log2 fold cKO –10

–10 Xpo5 Dgcr8cKO DicercKO Xpo5cKO –10 –5 0 5 10 Cont Xpo5 mean expression (Log)2

e Ctrl cKO g 1.2 #1 #2 11111 1 M Ctrl cKO Ctrl cKO

Cq) 0.8 ** 0.48

ΔΔ *** 0.6 0.32 (nt)

0.4 *** 0.15 *** *** 0.10 0.11 __ level ( Relative RNA 0.2 100 __ 0 70 pre-mir-17 miR17-5p miR-18 miR-19a miR-20a miR-92a 1 1.2 1 0.9 40 __ f Ctrl cKO h Ctrl cKO __ 1.2 1.4 NS 30 11 1 1 1.2 NS

Cq) mir-17 Cq) 1 0.8 20 __ ΔΔ

*** ΔΔ 0.8 ( 0.6 0.37 ( *** *** 0.6 1 0.1 1 0.1 0.4 0.20 0.20 0.4 0.2 5S RNA Relative RNA level 0.2 Relative RNA 0 level 0 miR-15a miR-15b miR-16 miR-320 miR-484

Fig. 8 XPO5 is required for miRNA biogenesis. a Bioinformatic pipeline of quantitative small RNA-seq and data analysis. b Small RNAs between 20 and 23 nt showed strong depletion in XPO5 cKO skin samples. Small RNA reads from 18 to 27 nt were charted from small RNA cDNA libraries. c The depletion of miRNAs in XPO5 cKO skin is weaker than that in Dicer1 and Dgcr8 cKO skin. For each boxplot, the middle line is the median, the vertical line spans the data range, and the hinges are the first and third quartiles. A two-sided Mann–Whitney test was used for statistical tests. d Depletion of mature miRNA reads in XPO5 cKO skin is evident for highly expressed miRNAs (red coloured dots). e Depletion of mir-17-5p, mir-18, mir-19a, and mir-20a except mir-92a in XPO5 cKO skin is measured by qPCR. f Depletion of mir-15a, mir-15b, and mir-16 in XPO5 cKO skin is measured by qPCR. g Depletion of mir-17-5p is confirmed by northern blotting. h Unchanged expression of mir-320 and mir-484 in XPO5 cKO skin is measured by qPCR. Data shown are mean s.d. from three independent experiments. **P < 0.01; ***P < 0.001, n.s.—not significant by Student’s t test. Original data for e–h are provided in the Source Data file. assays provided evidence that XPO5 binds to these unprocessed extensive base-pairing regions beyond pre-miRNA hairpin. primary miRNAs in a RanGTP-independent manner (Figs. 2 Because RanGTP-dependent cargo association is characteristic and 4). Studies of the mir-19a precursors further suggest that of exportin-mediated nuclear export1,theRanGTP- the binding site of XPO5 to these primary miRNAs is in the independent association with these primary miRNAs suggests

NATURE COMMUNICATIONS | (2020) 11:1845 | https://doi.org/10.1038/s41467-020-15598-x | www.nature.com/naturecommunications 13 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15598-x a hitherto unrecognized function of XPO5. Indeed, incubation KO may not be entirely due to the loss of miRNA biogenesis. of XPO5 with both pri-mir-19a, pri-mir-15b~16-2,andpri-mir- This possibility warrants future investigations. 15b increased the processing efficiency of the DROSHA/ DGCR8 microprocessor, providing evidence for the involve- Methods ment of XPO5 in the nuclear cleavage of closely clustered, Mice. All WT and transgenic animal breeding and operation procedures were polycistronic miRNAs. Given the lack of association of XPO5 approved by the Institutional Animal Care and Use Committees (IACUC) at the with monocistronic or sparsely clustered miRNAs, it is possible University of Colorado Boulder and in accordance with the guidelines and reg- that the XPO5-mediated mechanism for efficient pri-miRNA ulations for the care and use of laboratory animals, and compiled with all relevant ethical regulations for animal testing and research, and received relevant ethical cleavage is unique to closely clustered miRNAs with extensive approvals. The XPO5KI animal was purchased as an ES cell clone from the EU’sKO dsRNA regions. It is tempting to speculate that closely clustered mouse consortium. EIIa-Cre mice were obtained from the Jackson laboratory (JAX miRNAs may use multiple mechanisms including XPO5 bind- #003724). The mouse was crossed with Flipase expressing mice to generate the fl fl fl fl fl fl ing to regulate their biogenesis, conferring more complex XPO5 / mice. XPO5KI, XPO5 / , and K14-Cre; XPO5 / mice were bred and housed in the University of Colorado Boulder. Embryos studies were timed by the control to the production of these miRNAs post- presence of a plug, indicating gestational age E0.5. All mouse experiments were transcriptionally. However, how many of these closely clustered conducted in accordance with animal protocols approved by the IACUC. miRNAs whose nuclear cleavage is regulated by XPO5 should be investigated in future studies in a cell type-specific manner. Cell culture. HEK293T cells, obtained from ATCC (ATCC CRL-3216), were Second, XPO5 pervasively associates with many cellular RNAs cultured and maintained in DMEM (GIBCO) supplemented with 10% heat- in addition to miRNA precursors. Global RNA folding analyses inactivated fetal bovine serum and 1% penicillin/streptomycin (GIBCO) in the 5% fl/fl fl/fl and individual case examination suggest that XPO5 has a pre- CO2 incubator at 37 °C. XPO5 MEF cells were isolated from E14 XPO5 mice ference to dsRNA regions. In addition to miRNA precursors, using the Pierce™ Mouse Embryonic Fibroblast Isolation Kit (#88279, Thermo Fisher Scientific). XPO5 KO MEF cells were generated by Adeno-cre virus infec- XPO5 binds numerous cellular RNAs such as vtRNA, 7SL RNA, tion. XPO5fl/fl and XPO5 KO MEF cells were cultured and maintained in DMEM snRNA, snoRNA, 7SK RNA, tRNA, hTR, and SINE and LINE for Pierce™ Primary Cell Isolation Kits (#88287, Thermo Fisher Scientific) sup- repeat elements, which have diverse cellular localization and plemented with 10% heat-inactivated fetal bovine serum and 1% penicillin/strep- functions42. Interestingly, the Kd of XPO5 to pri-mir-19a variants tomycin (GIBCO) in the 5% CO2 incubator at 37 °C. and vtRNA is similar to the Kd of MDA5 and dsRNA43. Although functional consequences of these binding events will be RNA purification and mRNA and miRNA qPCR. Total RNA was extracted using examined in future studies, it is possible that XPO5 has multiple Trizol (Thermo Fisher Scientific). For mRNA analysis, 1 μg total RNA was used to 1 synthesize cDNA by Superscript III Reverse Transcriptase (Thermo Fisher Scien- functions beyond miRNA biogenesis . We note mammalian fi fi ti c). For miRNA analysis, the miScript II RT Kit (Qiagen) was used to synthesize XPO5 was originally identi ed as a binding protein to dsRNA cDNA. Reactions were performed according to the manufacturer’s manual and on binding proteins such as ILF3, PKR, and STAUFEN16. Our a CFX384 real-time system (Bio-Rad). Differences between samples and controls results now provide a molecular basis at the genomic scale for were calculated using the 2−ΔΔC(t) method. All the primers used for qPCR are these earlier findings and establish a foundation to investigate the listed in Supplementary Table 1. link between XPO5 and cellular dsRNAs and their binding proteins. Immunostaining and images. For analysis of back skin phenotypes, OCT sections Finally, we have unequivocally demonstrated the require- were fixed in 4% PFA for 10 min in phosphate-buffered saline (PBS) and washed three times for 5 min in PBS at room temperature. Block the sections with 2.5% ment of XPO5 for global miRNA biogenesis and mouse devel- NGS, 2.5% NDS in PBS. Primary antibodies against the following proteins were opment by examining XPO5 KO in embryonic and skin used: β4-integrin (β4, 1:100, BD Biosciences), Lef1 (1:500, Cell signaling), K5 development. In both systems examined, we observed strong (1:5000, Covance), K1 (1:2000, Covance), and Ki67 (1:500; Abcam). Imaging was developmental defects reminiscent of those observed in Dicer1 performed on a Leica DM5500B microscope with an attached Hamamatsu and Dgcr8 KO animals. In the skin, we have previously shown C10600-10B camera and MetaMorph (version 7.7; MDS Analytical Technologies) software. For all images, single optical sections were used. that the loss of Dicer1 or Dgcr8 leads to complete depletion of most miRNAs with a couple of exceptions such as mir-320 and X-gal histochemistry staining. Thick sections (25 µm) of back skin samples were mir-484,whichareDicer1-dependent but Dgcr8-indepen- fi fi 18,19 xed in x solution (0.5% glutaraldehyde, 1.25 mM EGTA pH 7.3, 2 mM MgCl2, dent . These cKO animals were neonatal lethal and died 1xPBS) for 10 min at room temperature. For the embryos, whole-mount embryos before one-week old. In contrast, KO of Ago1 and Ago2 together were fixed in fix solution for 5 min at room temperature. Fixed sections and (dKO) leads to 70–80% depletion of most miRNAs due to embryos were washed twice with detergent rinse (100 mM phosphate buffer pH quantitative loss of AGO proteins and the dKO mice usually 7.4, 2 mM MgCl2, 0.01% sodium deoxycholate, 0.02% NP-40), and then incubated 12 in staining solution (100 mM phosphate buffer pH 7.4, 2 mM MgCl2, 0.02% NP-40, survived to adulthood with prominent defects in hair follicles . 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, 20 mM Tris pH 7.5, Collectively, these models set useful boundaries to estimate the 1 mg/mL X-gal) at 37 °C for 48 h. requirement of XPO5 for miRNA biogenesis if miRNAs are not completely depleted in the absence of XPO5.OurmiR-seq, HITS-CLIP. XPO5 HITS-CLIP and data analyses were performed as described44 qPCR and northern blotting analyses suggested that KO of with slight modifications41. In Brief, HEK293T cells were UV-crosslinked, resus- XPO5 leads to ~90% depletion of most miRNAs. Of note, pended in lysis buffer. Supernatants were incubated with Anti-Exportin-5 antibody Dgcr8-independent miRNAs, mir-320 and mir-484,werenot (Abcam) or control IgG for 1 h on ice with or without RNase. After incubation, dynabeads were added into the protein–RNA complexes solutions and incubate for affected by the loss of XPO5. This result is consistent with a ′ γ 9 1 h at 4 °C, followed by labeling 5 end of RNA with [ -32P] ATP using PNK. Then previous report for XPO5-independent miRNA export and 5′ Ligation Adapter-NN was linked to 5′ end of RNA using T4 RNA ligase 1 at indicates an intrinsic link between DROSHA/DGCR8 produced room temperature for 2 h. Protein–RNA complexes were incubated in 1x Nu- pre-miRNA hairpin and XPO5-mediated export. In support of PAGE loading buffer (Invitrogen) at 70 °C for 10 min, resolved on 8% Nu-PAGE these molecular measurements, XPO5 skin cKO animals lived Bis-Tris gel (Invitrogen), transferred into a nitrocellulose membrane and exposed to a Fuji film. Bands corresponding to specific protein–RNA complexes were longer than their Dicer1 or Dgcr8 cKO counterparts but shorter excised, and then digested with proteinase K. RNAs were purified and adenylated than Ago1/2 dKO animals. Although other factors could weakly linker was linked to 3′ end of purified RNAs. cDNAs were made using Superscript compensate for the loss of XPO5 for miRNA expression, they III(Invitrogen) with 3′ adenylated linker primers. cDNAs were separated on 15% are unable to rescue the developmental defects. Therefore, we denaturing urea polyacrylamide gel and one area around 100 bp were excised purified, followed by amplification by Phusion High Fidelity polymerase (NEB). conclude that XPO5 is broadly required for miRNA biogenesis PCR products were separated by 6% denaturing urea polyacrylamide gel and three and mouse development. We note, however, because XPO5 also bands around 100–150 were purified. Obtained CLIP libraries were subjected to binds to many non-miRNA substrates the phenotypes of XPO5 high-throughput deep sequencing.

14 NATURE COMMUNICATIONS | (2020) 11:1845 | https://doi.org/10.1038/s41467-020-15598-x | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15598-x ARTICLE

Small RNA-seq. Total RNA samples from P2–P4 back skin epidermis were subject miR19a pri-miRNA substrates were incubated with XPO5 or XPO5 mutant protein to high efficiency 3′ ligation as described previously41. Products were resolved on at 25 °C for 25 min. Then, DROSHA and DGCR8 complex was added into system 15% PAGE-urea gels and stained with SybrGold (Molecular Probes S-11494 and incubated at 37 °C for 1 h in processing buffer (100 mM Tris-HCl pH 7.6, fi Carlsbad, CA, USA). The region corresponding to ligated miRNAs (46 nt) was 500 mM KCl, 1 mM EDTA, 64 mM MgCl2). Processing products were puri ed by excised, gel slices were thoroughly minced, and eluted in HSCB (400 mM NaCl, 25 phenol/chloroform and run on the 5% PAGE denaturing gel. Gels were dried and mM Tris-HCl pH 7.5, 0.1% SDS) overnight at 4 °C. Nucleic acids were precipitated detected by phosphorimager. in the presence of glycogen carrier via 0.1 volumes sodium acetate and 2.5 volumes ethanol. Pellets were washed in 70% ethanol, and dissolved in the 5′ ligation mix without enzyme. Samples were briefly heated to 70 °C, snap chilled on ice, and then RNA northern blot. Nuclear and cytoplasmic RNAs were extracted from MEF cells enzyme was added. Following ligation reactions cDNA was prepared using using the Cytoplasmic & Nuclear RNA Purification Kit (# 21000, Norgen Biotek). Superscript III RT (Invitrogen) according to the manufacturer′s recommendation Three micrograms of cytoplasmic RNA and two micrograms of nuclear RNA were using 3′ linker-specific RT primer. cDNA libraries served as templates for PCR separated by electrophoresis on 5% PAGE gels and electrically transferred into amplification; amplicons were resolved on 8% native acrylamide gels. Bands of the nylon N+ membrane. [γ-32P] ATP-labeled specific oligonucleotide probe correct molecular weight were isolated and used for high-throughput sequencing sequences (vtRNA and 5S) were used and hybridized at 42 °C overnight. Total on the Illumina HiSeq2000 and HiSeq4000. RNA was extracted from P4 back skin epidermis using Trizol (# 15596-018, Thermo Fisher Scientific). five micrograms of total RNA was separated by elec- trophoresis on 5% PAGE gels and electrically transferred into nylon N+ mem- Plasmids construction and transfection. Constructs coding for FLAG-tagged brane. [γ-32P] ATP-labelled specific oligonucleotide LNA probe sequences (mir- human XPO5, XPO5 mutant and RanQ69L were cloned in the pcDNA3.1 vector 17a) was used and hybridized at 42 °C overnight. Nonhybridized probes were for transient expression. For the mir-17~92a pri-miRNA in vitro transcription washed away by wash buffer (2x SSC/0.2%SDS) for 20 min twice. Signals were plasmid pJMC101-T7-pri-miR17~92a, human mir-17~92a pri-miRNA was cloned detected by phosphorimager. using mir-17~92a pri-miRNA transcription forward and reverse primers (includ- ing T7 promotor) and T7:pri-mir-17~92a were inserted into pJMC101vector. Primers for plasmid construction are listed in Supplementary Table 1. HEK293FT ® CLIP analysis. FASTQ files were trimmed using cutadapt (v1.8.3) (3′ adapter, transfection was performed using Mirus TransIT -LT1 Transfection Reagent (# TGGAATTCTCGGGTG CCAAG G, 5′ adapter CTACAGTCCGACGATC) to MIR 2300, Fisher) according to the manufacturer’s instructions. remove adapter sequences. The randomized 5′ and 3′ NN bases of the adapter were next appended to the FASTQ read ID. Reads were mapped to the human genome fi Protein expression and puri cation. For XPO5-His6 and His6-RanQ69L proteins (hg19 build) using novoalign (v3.02.00). Reads > 25 nt after trimming were aligned expression, pQE60-XPO5 and pQE32-RanQ69L plasmids are transformed sepa- with -l 25 -t 85 settings, whereas shorter reads were aligned with -l 20 -t 30 settings. rately into E.coli strain TG1 cells16. The XPO5 subcloned cells were grown in LB PCR duplicates were removed using UMI-Tools (v. 0.5.3) with the “directional” medium with 2% ethanol (vol/vol) at 37 °C, followed by induction with 400 µM option and alignments with MAPQ < 10 were discarded. Peaks were called by fi IPTG at 18 °C for 16 h. XPO5-His6 was puri ed on NTA-Ni2 beads (QIAGEN). merging all overlapping alignments using bedtools merge (v2.26.0). Peaks were For His6-RanQ69L protein, subcloned cells were grown in LB medium at 37 °C, annotated based on overlaps with annotations downloaded from the UCSC table μ followed by induction with 100 M IPTG at room temperature overnight. His6- browser (hg19). Peaks were intersected with the following annotations in order RanQ69L was purified on NTA-Ni2 beads and loaded GTP on ice for 3 h29. requiring 1 bp overlap, miRNA, tRNA, snoRNA, lincRNA, utr3, utr5, cds, ncRNA, repeatMasker, introns, mitochondria, retroelements, pseudogenes, or intergenic. Peaks were classified based on the first intersecting annotation. Repeat associated Electrophoretic mobility shift assay . Cold mir-17~92a pri-miRNA, pre-mir-30a, regions/RNA annotations are peaks that intersect with repeatMasker annotations pre-mir-19a, pri-mir-19a, pri-mir-19a truncation v1, pri-mir-19a truncation v2, from UCSC genome browser. Peaks were filtered to only keep peaks present in pri-mir-15b~16-2, pri-mir-15b, and vtRNA1-1 transcripts were produced by the fi ™ fi three of ve libraries, and with a minimum read count of 5. Libraries were also MEGAshortscript T7 Transcription Kit (#AM1354, Thermo Fisher Scienti c) independently mapped to a database of pre-microRNA sequences (mirBase release using pJMC101-T7-pri-mir-17~92 template (digested by HindIII), T7-pre-mir-30a, 19) using blastn (v.2.2.28) to assess alignments with hairpin sequences. Publically T7-pre-mir-19a, T7-pri-mir-19a, T7-pri-mir-19a truncation v1, T7-pri-mir-19a available CLIP datasets were downloaded from GEO or ENCODE using the fol- truncation v2, T7-pri-mir-15b~16-2, T7-pri-mir-15b, and T7-vtRNA1-1 PCR lowing accessions: DICER1 (GSE55333), DROSHA and DGCR8 (GSE61979), and products separately as the templates (primers are listed in Supplementary Table 1), DGCR8 (ENCFF491LUG). and purified by 5% PAGE gel. The 5′-phosphote of those cold substrates were removed by Antarctic Phosphatase (# M0289S, NEB). After dephosphorylation, those substrates were end-labeled by [r-32P]-ATP using T4 Polynucleotide Kinase Small RNA analysis. FASTQ files from DGCR8cko and DICER1cko experiments (# M0201S, NEB). miR17~92a, miR19a, and miR15b-16-2 pri-miRNAs were were trimmed to remove 3′ adapter sequence (TCGTATGCCGTCTTCTGCTTG)19. refolded in refolding buffer {1XTHE (66 mM HEPES, 33 mM Tris, 1 mM EDTA), FASTQ files from XPO5cko experiments were trimmed to remove 3′ adapter 100 mM NH OAc, 5 mM MgOAc, 0.5% NP-40, 0.1 mM EDTA} at 60 °C for 10 4 sequence (TGGAATTCTCGG GTGCCAAGG), 5′ adapter (CTA- min and slowly cooled down to 25 °C in 25 min. After refolding, mir-17~92a, mir- CAGTCCGACGATC), and randomized NN nucleotides introduced by the adap- 19a, mir-15b, mir-15b~16-2 pri-miRNAs, and vtRNA1-1 were separately incubated ters from both the 5′ and 3′ end. Following trimming reads were then aligned with with XPO5 protein in a total volume of 10 μl at 25 °C for 45 min. Pre-mir-30a and Bowtie2 (v.2.1.0) to the mouse genome (mm10) using local alignment. A custom pre-mir-19a were incubated with XPO5 protein and RanQ69L. The binding buffer annotation file was built containing GENCODE annotations, v19 for human, M8 contained 20 mM HEPES (pH 7.3), 150 mM potassium acetate, 2 mM magnesium for mouse with Mt_rRNA, lincRNA, misc_RNA, rRNA, sRNA, sm_RNA, snRNA, acetate, 0.05% NP-40, 7 mM 2-mercaptoethanol, 1.5 μg/mL poly dIdC, and 0.2% snoRNA, vaultRNA, and tRNA biotypes, and miRBase microRNA annotations BSA. After incubation, 0.3 μg of heparin was added to the reaction and incubated (release 19). Reads overlapping annotations were counted using Htseq-count for another 5 min4. Samples were then analyzed by electrophoresis on a 0.7% (v.0.6.0), requiring a minimum MAPQ of 10. FASTQ files from HEK293T cells nondenaturing agarose gel (for mir-17~92a pri-miRNA) at 4 °C or 5% non- were trimmed to remove 3′ adapter sequence (TGGAATTCTCGGGTGCCAAGG), denaturing PAGE gel (for pre-mir-30a, pre-mir-19a, pri-mir-19a, pri-mir-15b, 5′ adapter (CTACAGTCCGACG ATC), and randomized NN nucleotides intro- mir-15b~16-2 pri-miRNAs, and vtRNA1-1) at room temperature. Gels were dried duced by the adapters from both the 5′ and 3′ end and aligned using Bowtie2 as and detected by phosphorimager. described above. For comparing XPO5 HITS-CLIP to small RNA-Seq, small RNA- Seq reads were aligned with instead with novoalign (-l 18 -h 100 -t 60 -n 99) and In vitro analysis of pri-miRNA processing.[α-32P]-CTP (#BLU008H250Uc, reads were enumerated by counting intersections with a pre-miRNA database from Perkin) labeled miR19a, miR15b, and miR15b-16-2 pri-miRNA transcripts were miRbase. produced by the MEGAshortscript™ T7 Transcription Kit (#AM1354, Thermo Fisher Scientific) using T7 promotor-driven pri-MIRNA gene PCR products as the templates, and purified by 5% PAGE gel. pCK-Drosha-FLAG and pCK-FLAG- RNA secondary structure analysis. RNA folding predictions for Fig. 5j were Dgcr8, pCDNA3-FLAG-XPO5, pCDNA3-FLAG-XPO5_mutant, and pCDNA3.1 generated using RNAFold web server (v2.3.3). Peaks > 200 nt in length were plasmids were separately transfected into Human 293FT cells. After 48 h of cul- excluded from the folding analysis. miRNA metagene plots were generated using turing, cells were harvested and opened by hypotonic gentle lysis buffer (10 mM the HTSeq python library (v0.6.0). The secondary structure of hTR in Fig. 5h was Tris-HCl pH 7.6, 10 mM NaCl, 2 mM EDTA, 0.5% Triton X-100, proteinase adapted from the telomerase database34,45. The secondary structures of noncoding inhibitor). DROSHA and DGCR8 complex, XPO5 protein, and XPO5 mutant RNAs in Figs. S2, S4, and S5 were generated using RNAFold web server (http://rna. protein were purified by ANTI-FLAG®M2 Affinity Gel (#A2220, Sigma). XPO5 tbi.univie.ac.at/cgi-bin/RNAWebSuite/RNAfold.cgi). and XPO5 mutant proteins were eluted from ANTI-FLAG® M2 Affinity Gel by adding excess 3X FLAG® Peptide (#F4799, Millipore Sigma). [α-32P]-CTP labeled miR15b-16-2, miR15b, and miR19a pri-miRNA substrates were refolded in Statistical information. For all experiments with error bars, the standard deviation refolding buffer (1XTHE (66 mM HEPES, 33 mM Tris, 1 mM EDTA), 100 mM was calculated to indicate the variation within each experiment. Numbers of ani- NH40Ac, 5 mM MgOAc, 0.5% NP-40, 0.1 mM EDTA) at 60 °C for 10 min and mals used for phenotype study has indicated in figures. Student’s t test was used for slowly cooled down to 25 °C in 25 min. After refolding, miR15b-16-2, miR15b, and most experiments. A two-sided Mann–Whitney test was used for Fig. 5j.

NATURE COMMUNICATIONS | (2020) 11:1845 | https://doi.org/10.1038/s41467-020-15598-x | www.nature.com/naturecommunications 15 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15598-x

Reporting summary. Further information on research design is available in 26. Chang, T.-C., Pertea, M., Lee, S., Salzberg, S. L. & Mendell, J. T. Genome-wide the Nature Research Reporting Summary linked to this article. annotation of microRNA primary transcript structures reveals novel regulatory mechanisms. Genome Res. 25, 1401–1409 (2015). Data availability 27. Du, P., Wang, L., Sliz, P. & Gregory, R. I. A biogenesis step upstream of microprocessor controls miR-17∼92 expression. Cell 162, 885–899 (2015). XPO5 HITS-CLIP data and small RNA-seq data are deposited in the Gene Expression 28. Han, J. et al. Posttranscriptional crossregulation between Drosha and DGCR8. Omnibus data repository with accession number GSE111964. The source data underlying Cell 136,75–84 (2009). Figs. 1a, 2a–e, g, i, 3a–c, 4b, d, f, g, 5c, f, 6f–h, 7d–g, l, and 8e–h and Supplementary 29. Bischoff, F. R., Klebe, C., Kretschmer, J., Wittinghofer, A. & Ponstingl, H. Figs. 2b, 3, 5b and 5c are provided as a Source Data file. All data that support the findings RanGAP1 induces GTPase activity of nuclear Ras-related Ran. Proc. Natl of this study are available from the corresponding author upon reasonable request. Acad. Sci. USA 91, 2587–2591 (1994). 30. Lee, Y. et al. The nuclear RNase III Drosha initiates microRNA processing. Received: 2 September 2019; Accepted: 17 March 2020; Nature 425, 415–419 (2003). 31. Gregory, R. I. et al. The microprocessor complex mediates the genesis of microRNAs. Nature 432, 235–240 (2004). 32. Zeng, Y. & Cullen, B. R. Structural requirements for pre-microRNA binding and nuclear export by Exportin 5. Nucleic Acids Res. 32, 4776–4785 (2004). 33. Zhang, Q., Kim, N.-K. & Feigon, J. Architecture of human telomerase RNA. References Proc. Natl Acad. Sci. USA 108, 20325–20332 (2011). 1. Güttler, T. & Görlich, D. Ran‐dependent nuclear export mediators: a 34. Chen, J.-L., Blasco, M. A. & Greider, C. W. Secondary structure of vertebrate structural perspective. EMBO J. 30, 3457–3474 (2011). telomerase RNA. Cell 100, 503–514 (2000). 2. Bohnsack, M. T., Czaplinski, K. & Gorlich, D. Exportin 5 is a RanGTP- 35. Gundelfinger, E. D., Di Carlo, M., Zopf, D. & Melli, M. Structure and dependent dsRNA-binding protein that mediates nuclear export of pre- evolution of the 7SL RNA component of the signal recognition particle. miRNAs. RNA 10, 185–191 (2004). EMBO J. 3, 2325–2332 (1984). 3. Lund, E., Güttinger, S., Calado, A., Dahlberg, J. E. & Kutay, U. Nuclear export 36. Kuglstatter, A., Oubridge, C. & Nagai, K. Induced structural changes of 7SL of microRNA precursors. Science 303,95–98 (2004). RNA during the assembly of human signal recognition particle. Nat. Struct. 4. Yi, R., Qin, Y., Macara, I. G. & Cullen, B. R. Exportin-5 mediates the nuclear Mol. Biol. 9, 740–744 (2002). export of pre-microRNAs and short hairpin RNAs. Genes Dev. 17, 3011–3016 37. Skarnes, W. C. et al. A conditional knockout resource for the genome-wide (2003). study of mouse gene function. Nature 474, 337–342 (2011). 5. Gwizdek, C. et al. Exportin-5 mediates nuclear export of minihelix-containing 38. Murchison, E. P., Partridge, J. F., Tam, O. H., Cheloufi, S. & Hannon, G. J. RNAs. J. Biol. Chem. 278, 5505–5508 (2003). Characterization of dicer-deficient murine embryonic stem cells. Proc. Natl 6. Okada, C. et al. A high-resolution structure of the pre-microRNA nuclear Acad. Sci. USA 102, 12135–12140 (2005). export machinery. Science 326, 1275–1279 (2009). 39. Wang, Y., Medvid, R., Melton, C., Jaenisch, R. & Blelloch, R. DGCR8 is 7. Ha, M. & Kim, V. N. Regulation of microRNA biogenesis. Nat. Rev. Mol. Cell essential for microRNA biogenesis and silencing of embryonic stem cell self- Biol. 15, 509–524 (2014). renewal. Nat. Genet. 39, 380–385 (2007). 8. Martinez, I. et al. An Exportin-1-dependent microRNA biogenesis pathway 40. Smibert, P., Yang, J.-S., Azzam, G., Liu, J.-L. & Lai, E. C. Homeostatic control during human cell quiescence. Proc. Natl Acad. Sci. USA 114, E4961–E4970 of Argonaute stability by microRNA availability. Nat. Struct. Mol. Biol. 20, (2017). 789–795 (2013). 9. Xie, M. et al. Mammalian 5′-capped microRNA precursors that generate a 41. Zhang, Z., Lee, J. E., Riemondy, K., Anderson, E. M. & Yi, R. High-efficiency single microRNA. Cell 155, 1568–1580 (2013). RNA cloning enables accurate quantification of miRNA expression by deep 10. Kim, Y.-K., Kim, B. & Kim, V. N. Re-evaluation of the roles of DROSHA, sequencing. Genome Biol. 14, R109 (2013). Export in 5, and DICER in microRNA biogenesis. Proc. Natl Acad. Sci. USA 42. Cech, T. R. & Steitz, J. A. The noncoding RNA revolution-trashing old rules to 113, E1881–E1889 (2016). forge new ones. Cell 157,77–94 (2014). 11. Golden, R. J. et al. An Argonaute phosphorylation cycle promotes microRNA- 43. Peisley, A. et al. Cooperative assembly and dynamic disassembly of MDA5 mediated silencing. Nature 542, 197–202 (2017). filaments for viral dsRNA recognition. Proc. Natl Acad. Sci. USA 108, 12. Wang, D. et al. Quantitative functions of Argonaute proteins in mammalian 21010–21015 (2011). development. Genes Dev. 26, 693–704 (2012). 44. Moore, M. J. et al. Mapping Argonaute and conventional RNA-binding 13. Grimm, D. et al. Fatality in mice due to oversaturation of cellular microRNA/ protein interactions with RNA at single-nucleotide resolution using HITS- short hairpin RNA pathways. Nature 441, 537–541 (2006). CLIP and CIMS analysis. Nat. Protoc. 9, 263–293 (2014). 14. Lu, S. & Cullen, B. R. Adenovirus VA1 noncoding RNA can inhibit 45. Podlevsky, J. D., Bley, C. J., Omana, R. V., Qi, X. & Chen, J. J.-L. The small interfering RNA and MicroRNA biogenesis. J. Virol. 78, 12868–12876 telomerase database. Nucleic Acids Res. 36, D339–D343 (2008). (2004). 15. Calado, A., Treichel, N., Müller, E.-C., Otto, A. & Kutay, U. Exportin-5- mediated nuclear export of eukaryotic elongation factor 1A and tRNA. EMBO Acknowledgements – J. 21, 6216 6224 (2002). We thank all members of the Yi laboratory for suggestions and helpful discussions, N. 16. Brownawell, A. M. & Macara, I. G. Exportin-5, a novel karyopherin, mediates Pace for generous support. We thank I. Macara for XPO5 and RanQ69L expression nuclear export of double-stranded RNA binding proteins. J. Cell Biol. 156, plasmids (Vanderbilt University School of Medicine), V. N. Kim for DROSHA and – 53 64 (2002). DGCR8 expression plasmids (Seoul National University), B. Gao and K. Diener for 17. Andl, T. et al. The miRNA-processing enzyme dicer is essential for the sequencing. Research reported in this publication was supported by the National Institute – morphogenesis and maintenance of hair follicles. Curr. Biol. 16, 1041 1049 of Arthritis and Musculoskeletal and Skin Diseases of the National Institutes of Health (2006). under Award Number R01AR059697 and R01AR066703 (to R.Y.). The content is solely 18. Yi, R. et al. Morphogenesis in skin is governed by discrete sets of differentially the responsibility of the authors and does not necessarily represent the official views of – expressed microRNAs. Nat. Genet. 38, 356 362 (2006). the National Institutes of Health. J.E.L. was supported by an American Cancer Society 19. Yi, R. et al. DGCR8-dependent microRNA biogenesis is essential for skin postdoctoral fellowship (#125209). Y.Y. was supported by an NYSTEM postdoctoral – development. Proc. Natl Acad. Sci. USA 106, 498 502 (2009). fellowship. Work in E.C.L.’s group was supported by the National Institutes of Health fi 20. Seong, Y. et al. Global identi cation of target recognition and cleavage by (R01-GM083300 and R01-HL135564) and MSK Core Grant P30-CA008748. the microprocessor in human ES cells. Nucleic Acids Res 42, 12806–12821 (2014). 21. Van Nostrand, E. L. et al. Robust transcriptome-wide discovery of RNA- Author contributions binding protein binding sites with enhanced CLIP (eCLIP). Nat. Methods 13, R.Y. conceived the study. J.W., J.E.L., Y.Y., E.C.L., and R.Y. designed the experiments. – 508 514 (2016). J.W. performed embryonic and skin studies with assistance from J.E.L. and XPO5 22. Rybak-Wolf, A. et al. A variety of dicer substrates in human and C. elegans. binding assays and pri-miRNA processing assays with assistance from S.M.M. J.E.L. – Cell 159, 1153 1167 (2014). established mouse models, performed initial molecular, and phenotypical studies, per- fi fi 23. Kozomara, A. & Grif ths-Jones, S. miRBase: annotating high con dence formed XPO5 HITS-CLIP experiments. K.R. performed bioinformatics analysis related to – microRNAs using deep sequencing data. Nucleic Acids Res. 42, D68 D73 HITS-CLIP and miR-seq. R.Y. supervised the study, analyzed the data, and wrote the (2014). manuscript with input from all authors. 24. Nguyen, T. A. et al. Functional anatomy of the human microprocessor. Cell 161, 1374–1387 (2015). 25. Kim, V. N., Han, J. & Siomi, M. C. Biogenesis of small RNAs in animals. Nat. Competing interests Rev. Mol. Cell Biol. 10, 126–139 (2009). The authors declare no competing interests.

16 NATURE COMMUNICATIONS | (2020) 11:1845 | https://doi.org/10.1038/s41467-020-15598-x | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15598-x ARTICLE

Additional information Open Access This article is licensed under a Creative Commons Supplementary information is available for this paper at https://doi.org/10.1038/s41467- Attribution 4.0 International License, which permits use, sharing, 020-15598-x. adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Correspondence and requests for materials should be addressed to R.Y. Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless Peer review information Nature Communications thanks Andor Pivarcsi and the other, indicated otherwise in a credit line to the material. If material is not included in the anonymous, reviewer(s) for their contribution to the peer review of this work. Peer article’s Creative Commons license and your intended use is not permitted by statutory reviewer reports are available. regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ Reprints and permission information is available at http://www.nature.com/reprints licenses/by/4.0/.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. © The Author(s) 2020

NATURE COMMUNICATIONS | (2020) 11:1845 | https://doi.org/10.1038/s41467-020-15598-x | www.nature.com/naturecommunications 17